[go: up one dir, main page]

WO2018181094A1 - Système d'aide à la traduction et analogue - Google Patents

Système d'aide à la traduction et analogue Download PDF

Info

Publication number
WO2018181094A1
WO2018181094A1 PCT/JP2018/011977 JP2018011977W WO2018181094A1 WO 2018181094 A1 WO2018181094 A1 WO 2018181094A1 JP 2018011977 W JP2018011977 W JP 2018011977W WO 2018181094 A1 WO2018181094 A1 WO 2018181094A1
Authority
WO
WIPO (PCT)
Prior art keywords
source language
language sentence
machine
input
translation
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Ceased
Application number
PCT/JP2018/011977
Other languages
English (en)
Japanese (ja)
Inventor
修一 倉林
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Cygames Inc
Original Assignee
Cygames Inc
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Cygames Inc filed Critical Cygames Inc
Priority to CN201880035286.3A priority Critical patent/CN110678868B/zh
Publication of WO2018181094A1 publication Critical patent/WO2018181094A1/fr
Priority to US16/586,000 priority patent/US11288460B2/en
Anticipated expiration legal-status Critical
Ceased legal-status Critical Current

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F40/00—Handling natural language data
    • G06F40/40—Processing or translation of natural language
    • G06F40/58—Use of machine translation, e.g. for multi-lingual retrieval, for server-side translation for client devices or for real-time translation
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F40/00—Handling natural language data
    • G06F40/40—Processing or translation of natural language
    • G06F40/42—Data-driven translation
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F40/00—Handling natural language data
    • G06F40/40—Processing or translation of natural language
    • G06F40/51—Translation evaluation

Definitions

  • the present invention relates to a translation support system, and more particularly to a translation support system, a translation support apparatus, a program, and a method for supporting machine translation from a source language sentence to a target language sentence.
  • machine translation engines machine translation engines
  • RBMT Rule Base Machine Translation
  • SMT statistical machine translation
  • Rule-based machine translation takes a method of parsing the source language, translating it into the target language using a dictionary, and rearranging the translated character strings into the target language grammar.
  • Rule-based machine translation uses knowledge prepared manually, such as language rules, grammatical rules, and dictionaries, in syntactic analysis and phrase translation.
  • rule-based machine translation is realized by a human programming a translation dictionary and translation rules. At that time, the programmer is familiar with the grammar of both the source language and the target language, thinks about rules for changing the word order or replacing with corresponding words, and writes down the rules in a form that can be executed as a program.
  • Statistical machine translation is an approach that automatically generates a translation model from a large number of parallel translations (corpus) (Non-Patent Document 1).
  • Typical implementations include Google's Translate API and Microsoft's Translate API.
  • Statistical machine translation can automatically generate language rules, grammar rules, and dictionaries required by rule-based machine translation from a corpus, so that a translation system can be constructed at low cost.
  • statistical machine translation can perform feedback learning by selecting appropriate results from a plurality of translation result candidates according to feedback from users, and can gradually improve translation accuracy. There are advantages. In this feedback learning, when a user selects or inputs a more appropriate translation sentence, the input can be used as feedback, and an appropriate result can be automatically selected from a plurality of parallel translation result candidates from the next time. . Due to the above features, statistical machine translation is extremely useful in translations targeting multiple languages (for example, more than several tens of languages).
  • the rule-based machine translation has a stable and high accuracy in the translation of various sentences, but it takes time and effort to manually construct language rules, grammar rules, and dictionaries.
  • these language rules, grammatical rules, and dictionaries need to be created for both the source language and the target language, and the cost of performing translation for multiple languages (for example, several tens of languages or more) is further increased. There is a tendency.
  • Statistic machine translation is difficult to apply to fields where a large amount of corpora cannot be prepared, and difficult to apply to fields where a large number of unknown words can occur. Also, since the translation accuracy depends on the corpus, it is difficult to keep the translation accuracy constant. For example, when statistical machine translation is applied to the game field, it is necessary to prepare and learn a large number of parallel translations of sentences close to a story to be translated. There are many narrative texts in games (scenarios such as in-game scenarios) that contain unusual expressions and ambiguous text whose meaning depends on the information behind the game. Machine translation is difficult. Although feedback learning is suitable for web translation with a large number of users, it cannot do multilingual learning at the same time. It is difficult to apply.
  • the method of machine translation after rewriting a source language sentence to a CNL sentence once is to obtain accurate translation results in multiple languages by rewriting all the original sentences to CNL sentences (sentences that satisfy certain CNL rules).
  • CNL sentences sentences that satisfy certain CNL rules.
  • CNL is not suitable for expressing the subtle nuances inherent in natural language, and it may not always be appropriate to rewrite all sentences into CNL sentences.
  • the high accuracy of translation means the accuracy of translation including maintaining the nuance of the original text.
  • the present invention has been made to solve such problems, and reduces translation costs by reducing the involvement of humans while maintaining high translation accuracy using a machine translation system.
  • the main purpose is to provide a translation support system that can do this.
  • a translation support system for supporting machine translation from a source language sentence to a target language sentence.
  • An input unit that accepts input, an error database that stores at least words or combinations of words contained in multiple source language sentences that are not correctly translated from the source language sentence to the target language sentence, and multiple source language sentences , and a controlled source language sentence database that stores controlled source language sentences that are controlled source language sentences expressed in a format that satisfies a predetermined condition corresponding to the plurality of source language sentences, and the input source
  • a control unit that determines whether or not a language sentence is machine-translatable, and an output unit that can output the input source language sentence that is determined to be non-machine-translatable.
  • a score representing the complexity of the source language sentence is determined based on the sentence structure of the input source language sentence, and whether or not the input source language sentence is machine-translatable based on the magnitude of the score. If it is determined that machine translation is possible, the input based on the word or combination of words included in the input source language sentence and the word or combination of words stored in the error database A word or a word included in the input source language sentence when it is further determined whether the input source language sentence is machine-translatable and it is determined that the input source language sentence is not machine-translatable Is compared with a word or a combination of words contained in a source language sentence stored in the controlled source language sentence database, thereby converting the input source language sentence into the controlled source language sentence. If the input source language sentence is converted to the controlled source language sentence, the input source language sentence is determined to be machine-translatable and not convertible. Is determined, it is determined that the input source language sentence is not machine-translatable.
  • control unit includes a score representing the complexity of the input source language sentence, the length of the source language sentence, the number of predetermined parts of speech included in the source language sentence, and The determination is based on at least one of the predetermined number of words included in the source language sentence.
  • the translation support system further includes a source language sentence corpus that stores a plurality of source language sentences, and each of the stored source language sentences has a score representing the complexity of the source language sentence.
  • the control unit associates a score representing the complexity of the input source language sentence with a score associated with the stored source language sentence having a high similarity to the input source language sentence. Determine based on.
  • the controlled source language sentence is a source language sentence expressed using grammar, syntax, and vocabulary suitable for machine translation.
  • the input unit can further accept an input of a controlled source language sentence for a source language sentence that is determined not to be convertible into the controlled source language sentence.
  • a source language that is determined not to be convertible to the controlled source language sentence is converted to the controlled source language sentence that has received the input and is determined not to be convertible to the controlled source language sentence.
  • the sentence and the controlled source language sentence that has received the input are stored in the controlled source language sentence database.
  • control unit uses a machine translation system to machine-translate the input source language sentence determined to be machine-translatable and the converted source language sentence converted. And the machine-translated target language sentence is output to the output unit.
  • the input unit can further accept that the machine-translated target language sentence is not properly translated, and the control unit corresponds to the machine corresponding to the target language sentence.
  • the input source language sentence before translation execution is stored in the error database.
  • control unit replaces a predetermined word included in the input source language sentence before execution of the machine translation with a predetermined proper noun, and outputs the machine translated target language sentence.
  • the substituted proper noun included is replaced again with the word before the replacement.
  • control unit periodically performs machine translation on a predetermined source language sentence using the machine translation system, the machine translated target language sentence, and the machine translation A change in the machine translation system is detected based on a comparison with a target language sentence that has been machine translated immediately before.
  • a translation support apparatus is a translation support apparatus for supporting machine translation from a source language sentence to a target language sentence, and is a translation target apparatus.
  • An input unit that accepts input of a language sentence, an error database that stores at least words or combinations of words included in a plurality of source language sentences in which machine translation from the source language sentence to the target language sentence is not performed correctly, and a plurality of source words
  • a controlled source language sentence database that stores a controlled source language sentence that is a controlled source language sentence expressed in a format that satisfies a predetermined condition corresponding to the language sentence and the plurality of source language sentences;
  • a control unit that determines whether or not the source language sentence is machine-translatable, and an output unit that can output the input source language sentence that is determined not to be machine-translatable.
  • a score representing the complexity of the source language sentence is determined based on the sentence structure of the input source language sentence, and whether or not the input source language sentence is machine-translatable is determined based on the magnitude of the score If it is determined that machine translation is possible, the input is performed based on the word or combination of words included in the input source language sentence and the word or combination of words stored in the error database. Whether or not the source language sentence is machine-translatable, and if it is determined that the input source language sentence is not machine-translatable, the words or words included in the input source language sentence The input source language sentence can be converted into the controlled source language sentence by comparing the combination with a word or a combination of words contained in the source language sentence stored in the controlled source language sentence database. If the input source language sentence is converted to the controlled source language sentence, it is determined that it can be machine-translated and cannot be converted. If it is determined, it is determined that the input source language sentence is not machine-translatable.
  • a program as one aspect of the present invention is a program for supporting machine translation from a source language sentence to a target language sentence, and the computer is configured to transmit the source language to be translated.
  • An input receiving step for receiving an input of a sentence, and a score representing the complexity of the input source language sentence is determined based on the sentence structure of the input source language sentence, and the input is performed according to the size of the score
  • a first determination step for determining whether or not the source language sentence is machine-translatable, and when it is determined that machine translation is possible in the first determination step, the source language sentence is included in the input source language sentence
  • a second determination step for determining whether or not the input source language sentence is machine-translatable based on a word or a combination of words and the first determination step or the second determination When it is determined in the step that machine translation is not possible, a word or a combination of words included in the input source language sentence, a plurality of source language sentences and a format that satisfies a predetermined condition corresponding to the plurality of source language sentences
  • a word or a combination of words included in a source language sentence stored in a database that stores a controlled source language sentence that is a controlled source language sentence represented by It is determined whether or not it can be converted into the controlled source language sentence.
  • the input source language sentence is converted into the controlled source language sentence and machine translation is possible.
  • a third determination step that determines that the input source language sentence is not machine-translatable and a determination that the machine-translation is not possible in the third determination step.
  • a method for supporting machine translation from a source language sentence to a target language sentence, wherein the source language sentence to be translated is input.
  • An input receiving step for receiving the input, a score representing the complexity of the input source language sentence is determined based on the sentence structure of the input source language sentence, and the input source language sentence is determined according to the magnitude of the score
  • a word or a word included in the input source language sentence when it is determined in the first determination step that the machine translation is possible and in the first determination step that the machine translation is possible
  • a word or combination of words stored in a database that contains at least a word or combination of words contained in multiple source language sentences in which machine translation from the source language sentence to the target language sentence is not performed correctly
  • a second determination step for determining whether or not the input source language sentence is machine-translatable based on the set, and machine translation is not possible in the first determination step or the second determination step.
  • a word or a combination of words included in the input source language sentence is controlled in a form that satisfies a plurality of source language sentences and a predetermined condition corresponding to the plurality of source language sentences.
  • the input source language sentence is converted into the controlled source language sentence by comparing with a word or a combination of words included in the source language sentence stored in the database storing the controlled source language sentence that is a source language sentence. If it is determined that conversion is possible, the input source language sentence is converted to the controlled source language sentence and determined to be machine-translatable. Possible
  • a third determination step that determines that the input source language sentence is not machine-translatable, and the input source language sentence that is determined not to be machine-translatable in the third determination step. And an output step for outputting.
  • FIG. 1 is an overall configuration diagram of a system according to an embodiment of the present invention. It is a block diagram which shows the hardware constitutions of the translation assistance apparatus by one Embodiment of this invention. An example of the functional block diagram of the translation assistance apparatus by one Embodiment of this invention is shown. It shows a state of filter processing by a character string conversion filter unit of a machine translation conversion unit according to an embodiment of the present invention. It is a flowchart which shows the process of the machine translation availability determination of the translation assistance apparatus by one Embodiment of this invention. It is a flowchart which shows the machine translation process of the translation assistance apparatus by one Embodiment of this invention. It is a whole block diagram of the translation assistance system by other one Embodiment of this invention.
  • a translation support system is a system that improves the accuracy of translation by controlling an input source language for an existing machine translation system, and has a one-level abstraction level than an existing machine translation system. Is a high meta-level translation system.
  • the translation support system according to the embodiment of the present invention performs multilingualization of a game that displays a large amount of text, represented by Japanese RPG, with high accuracy and low cost using an existing machine translation system. Is a meta-level translation system.
  • the source language sentence is a sentence composed of a language to be translated (translation source language), and the target language sentence is a sentence composed of a translated language (translation destination language).
  • the sentence includes incomplete sentences such as sentences with nouns only.
  • the source language is assumed to be English, but other languages may be used.
  • the description when referring to one or more sentences and each sentence is described in the subsequent description, the description also includes one sentence.
  • FIG. 1 is an overall configuration diagram of a translation system 2 according to an embodiment of the present invention.
  • the translation system 1 includes a translation support device 10 and a machine translation server 6.
  • the translation support apparatus 10 and the machine translation server 6 are connected to a network 4 such as the Internet and can communicate with each other. However, each may be individually connected as necessary.
  • the translation support system 1 has a configuration other than the machine translation server 6 in the translation system 2.
  • FIG. 2 is a block diagram showing a hardware configuration of the translation support apparatus 10 according to an embodiment of the present invention.
  • the translation support apparatus 10 includes the same configuration as a general server or PC.
  • the translation support device 10 includes a processor 11, an output device 12, an input device 13, a storage device 14, and a communication device 15. Each of these components is connected by a bus 16. It is assumed that an interface is interposed between the bus 16 and each component device as necessary.
  • the processor 11 controls the overall operation of the translation support apparatus 1 and is, for example, a CPU. Note that an electronic circuit such as an MPU may be used as the processor 11.
  • the processor 11 executes various processes by reading and executing programs and data stored in the storage device 14.
  • the input device 12 is a user interface that receives input from the user to the translation support device 10, and is, for example, a touch panel, a touch pad, a keyboard, or a mouse.
  • the output device 13 outputs or displays the output information of the translation support system 1 to the user, and is, for example, a display that outputs an image.
  • the output device 13 can also include a printer.
  • the storage device 14 includes a main storage device and an auxiliary storage device.
  • the main storage device is a semiconductor memory such as a RAM.
  • the RAM is a volatile storage medium capable of reading and writing information at high speed, and is used as a storage area and a work area when the processor 11 processes information.
  • the main storage device may include a ROM that is a read-only nonvolatile storage medium. In this case, the ROM stores a program such as firmware.
  • the auxiliary storage device stores various programs and data used by the processor 11 when executing each program.
  • the auxiliary storage device is, for example, a hard disk device, but may be any non-volatile storage or non-volatile memory as long as it can store information, and may be removable.
  • the auxiliary storage device stores, for example, an operating system (OS), middleware, application programs, various data that can be referred to when these programs are executed.
  • OS operating system
  • middleware middleware
  • application programs various data that can be referred to when these programs are executed.
  • the communication device 15 is a device for exchanging data with other computers via the network 4.
  • the communication device 15 performs wired communication using an Ethernet (registered trademark) cable, mobile communication, wireless communication such as a wireless LAN, and connects to the network 4.
  • Ethernet registered trademark
  • mobile communication such as a wireless LAN
  • the translation support apparatus 10 has a database server function.
  • the storage device 14 stores data (for example, tables) and programs for various databases, and various databases are realized by executing the programs.
  • the translation support apparatus 10 includes a plurality of computers (servers).
  • the translation support device 10 may include a database server.
  • the machine translation server 6 is a server that executes machine translation, and provides machine translation in response to access from any client terminal including the translation support apparatus 10.
  • the machine translation server 6 is an existing machine translation system such as Google Translate API or Microsoft Translate API.
  • the machine translation server 6 executes machine translation from the source language specified by the translation support apparatus 10 to the target language for the source language sentence received from the translation support apparatus 10, and translates the machine translated target language sentence into the translation support apparatus. Reply to 10.
  • FIG. 3 shows an example of a functional block diagram of the translation support apparatus 10 according to an embodiment of the present invention.
  • the translation support apparatus 10 includes a control unit 21, an input unit 22, an output unit 23, a source language sentence database (DB) 24, an error DB 25, a controlled source language sentence DB 26, and a translation target data DB 27.
  • these functions are realized by a program being executed by the processor 11. However, it can also be realized by hardware by configuring an electronic circuit or the like for realizing each function.
  • another part may have a part of one part (function).
  • arrows in the drawing indicate main information exchange between the respective units, but are not limited to these as long as the operation according to the embodiment of the present invention can be realized.
  • the input unit 22 is configured using the input device 12 and is configured to receive an input from the user to the translation support device 10.
  • the output unit 23 outputs data and signals to the user via the output device 13.
  • the input unit 22 receives input of one or more source language sentences to be translated. At this time, the input unit 22 accepts input of designation of the source language and designation of the target language. The designation of the target language can also be configured such that the input unit 22 accepts input before transmission of the machine translation server 6.
  • the output unit 23 displays on the display a screen that accepts an input to be translated, selection of a source language sentence, and selection of a target language.
  • the selection of the source language sentence and the selection of the target language are selection of one of a plurality of languages to be translated that are targeted by the translation support system 1 in advance.
  • the translation target language targeted by the translation support system 1 is preferably the same as the language supported by the existing machine translation system.
  • one or a plurality of source language sentences to be translated, which are received by the input unit 22, are referred to as input source language sentences.
  • the translation target data DB 27 stores input source language sentences, and stores a machine translation availability determination flag for each sentence.
  • the control unit 21 updates the determination flag in accordance with the machine translation availability determination process described below.
  • the determination flag is an example, and the data stored in the translation target data DB 27 may be data that can identify whether each sentence of the input source language sentence is machine-translatable (Machine-Translatability).
  • the error DB 25 stores one or more source language sentences in which machine translation from the source language sentence to the target language sentence is not correctly performed.
  • the source language sentence in which the machine translation from the source language sentence to the target language sentence is not correctly performed means that the source language sentence when the target language sentence as a result of the machine translation actually performed by the machine translation server 6 is determined by a human being. The meaning of the sentence is different or the meaning is unclear.
  • the error DB 25 is included in one or a plurality of source language sentences obtained by using a part or all of the stored source language sentences and in which machine translation from the source language sentence to the target language sentence is not correctly performed.
  • One or more combinations of words to be stored are stored.
  • the word combination is composed of a plurality of words.
  • the word combination is a phrase or idiom composed of a plurality of words, or a simple word string.
  • the combination of words stored in the error DB 25 is one in which machine translation is not correctly performed when each word has a configured word.
  • the error DB 25 stores one or more words included in one or more source language sentences in which machine translation from the source language sentence to the target language sentence is not correctly performed. In another example, the error DB 25 includes one or more source language sentences in which machine translation from the source language sentence to the target language sentence is not correctly performed, and one or more words or combinations of words included in the source language sentence. Store.
  • the controlled source language sentence DB 26 stores a plurality of source language sentences and a controlled source language sentence that is a controlled source language sentence expressed in a format that satisfies a predetermined condition corresponding to the plurality of source language sentences. .
  • the controlled source language sentence DB 26 is a corpus that stores, for each language (source language), a plurality of source language sentences and a plurality of controlled source language sentences corresponding to the plurality of source language sentences. is there.
  • a controlled source language sentence is a source language sentence expressed using grammar, syntax, and vocabulary suitable for machine translation.
  • the controlled source language is a natural language whose grammar and vocabulary are controlled so as to satisfy a predetermined requirement generally called CNL (Controlled Natural Language).
  • CNL Controlled Natural Language
  • the controlled source language is CNL.
  • the controlled source language sentence DB 26 stores, as a pair, a result obtained by manually rewriting the input source language sentence into the CNL sentence.
  • the source language sentence DB 24 includes a source language sentence corpus that stores a plurality of source language sentences. Each stored source language sentence is associated with a score representing the complexity of the source language sentence. The score is obtained using a known method (for example, William H. DuBay, The Principles of Readability, 25 August 2004, http://www.impact-information.com/impactinfo/readability02.pdf). And the number of words or combinations of words and related pronouns included in the source language sentence.
  • the score is based on at least one of a predetermined number of parts of speech such as relative pronouns and conjunctions included in the source language sentence, and a predetermined number of words included in the source language sentence that are easily mistranslated. Calculated. In another example, the score is calculated based on at least one of omission of the subject, predicate, and object of the source language sentence and whether or not it is a basic sentence type.
  • the control unit 21 includes a basic control unit 31, a machine translation availability determination unit 32, a controlled source language sentence conversion unit 36, a DB control unit 37, a machine translation conversion unit 38, a machine translation server monitoring unit 39, A machine learning unit 40.
  • the basic control unit 31 performs basic control of the entire translation support apparatus 10.
  • the basic control unit 31 is a software module for realizing various basic functions, and is an OS, middleware, or the like.
  • the DB control unit 37 receives instructions from each unit and refers to and stores information in the various DBs 24 to 27.
  • a process in which each unit directly performs a DB operation will be described, but it is understood that the process is preferably performed via the DB control unit 37.
  • the machine translation enable / disable determining unit 32 includes a syntax filter unit 33, a semantic filter unit 34, and a controlled source language sentence conversion enable / disable determining unit 35.
  • the input source language sentence is machine-translatable (Machine-Translatable). Judge whether it is a source language sentence. Thereby, it becomes possible to extract the source language sentence which cannot be made into machine translation object. Similarly, it is possible to extract a source language sentence that can be a machine translation target.
  • the syntax filter unit 33 determines, for each sentence of the input source language sentence, a score representing the complexity of the input source language sentence based on the sentence structure of the input source language sentence, and the input level is determined based on the magnitude of the score. It is determined whether each sentence of the source language sentence is machine-translatable.
  • the score representing the complexity of the source language sentence is a quantitative score calculated using the length of the source language sentence, the included words or combinations of words, the number of related pronouns, and the like.
  • the DB control unit 37 updates (stores) the machine translation availability determination flag for each input source language sentence.
  • the syntax filter unit 33 extracts a source language sentence having a high similarity to the input source language sentence using the source language sentence corpus. In one example, similarity is determined by comparing sentence length, number of relative pronouns, number of conjunctions, and number of predetermined words (predetermined words) for two sentences to be compared. . The syntax filter unit 33 determines whether or not the input source language sentence can be machine-translated based on the score associated with the extracted source language sentence. Note that a source language sentence that can be machine translated (a source language sentence that can be machine translated) is a target language corresponding to the meaning of the source language sentence when the source language sentence is machine translated by the machine translation server 6. A sentence, that is, a source language sentence from which an appropriately translated target language sentence can be obtained.
  • a source language sentence that is not machine-translatable is synonymous with a source language sentence in which machine translation is not performed correctly, and the meaning of the source language sentence when the source language sentence is machine-translated by the machine translation server 6 Is a source language sentence from which a target language sentence having a different meaning or a target language sentence having an unclear meaning can be obtained.
  • the syntax filter unit 33 sets in advance a threshold value of a score at which machine translation is not performed correctly, and compares the threshold value of the score with a score associated with the extracted source language sentence, whereby the input source language sentence is Determine whether or not translation is possible.
  • the threshold value is set based on a score of a source language sentence in which machine translation is correctly performed and a score of a source language sentence in which machine translation is not performed correctly.
  • the threshold is set using the minimum score of a plurality of source language sentences in which machine translation is not performed correctly.
  • an average value and a standard deviation of scores of a plurality of source language sentences in which machine translation is performed correctly and an average value and a standard deviation of scores in a plurality of source language sentences in which machine translation is not performed correctly are used The threshold value is set.
  • the syntax filter unit 33 determines that a source language sentence having a score of less than 0.8, for example, can be machine translated, and machine language translation of a source language sentence having a score of 0.8 or more is performed Determine that it is not possible.
  • the syntax filter unit 33 quantitatively determines the complexity of the input source language sentence as a score, and can be machine-translated according to the determined score size. It is determined whether or not. This makes it possible to determine whether machine translation is possible from the complexity of the sentence structure of the input source language sentence.
  • the syntax filter unit 33 calculates a score representing the complexity of the source language sentence using the above-described known method, and the input source language sentence can be machine-translated based on the magnitude of the score. It is determined whether or not there is. For example, in calculating or determining the score, the syntax filter unit 33 calculates a score based on the length of the source language sentence, the word or combination of words included in the source language sentence, and the number of related pronouns. In such a case, since the syntax filter unit 33 can calculate or determine the score without using the source language sentence corpus, the translation support apparatus 10 does not have to include the source language sentence DB 24.
  • the semantic filter unit 34 For each sentence of the input source language sentence, the semantic filter unit 34 selects the input source language based on the word or combination of words included in the input source language sentence and the word or combination of words stored in the error DB 25. It is determined whether each sentence of the sentence is machine-translatable. Preferably, the semantic filter unit 34 determines the source language sentence determined by the syntax filter unit 33 to be machine-translatable. In one example, at this time, the DB control unit 37 updates the machine translation availability determination flag for each input source language sentence.
  • the input unit 22 can accept an input that the machine-translated target language sentence has not been properly translated.
  • the DB control unit 37 stores the input source language sentence corresponding to the target language sentence that has received the input in the error DB 25.
  • the machine learning unit 40 applies a machine learning algorithm to a part or all of the source language sentence stored in the error DB 25 and learns a feature amount common to a sentence group that is not machine-translatable.
  • the machine learning unit 40 applies a machine learning algorithm to some or all of the source language sentences stored in the error DB 25, and the words of words included in one or more source language sentences that are not correctly translated.
  • the combination is extracted and stored in the error DB 25.
  • the machine learning unit 40 can use a known machine learning algorithm such as a random forest method or a method using a neural network.
  • the semantic filter unit 34 uses the error DB 25 in which the quality of the target language sentence actually machine-translated is fed back, so that the input source language sentence is converted into a machine. It is determined whether or not translation is possible. This makes it possible to determine whether machine translation is possible from the semantic complexity of the input source language sentence.
  • the semantic filter unit 34 determines that the input source language sentence is not machine-translatable, If the input source language sentence is not included, it is determined that the input source language sentence can be machine translated.
  • the semantic filter unit 34 is a module that measures the semantic complexity of the original text using the error DB 25.
  • the source language sentence (sentence) stored in the error DB 25 can be expressed as a bag of words (BoW) type vector.
  • W i indicates the number of times a word i appears in the sentence s
  • the error DB 25 is a set S of these sentence vectors s. s ⁇ S.
  • the binary classification period classifier (q) generated using the machine learning unit 40 can be defined as follows.
  • true indicates that the sentence vector q is determined to be machine-translatable, and false indicates that the sentence vector q is determined not to be machine-translatable.
  • the output unit 23 displays an input source language sentence determined not to be machine-translatable on a display or outputs to a printer in accordance with an input to the input unit 22 received by a user operation.
  • the controlled source language sentence conversion enable / disable determining unit 35 determines that the input source language sentence determined by the semantic filter unit 34 as not machine-translatable is a controlled source language sentence (CNL sentence) based on the controlled source language sentence DB 26. It is determined whether or not it is convertible.
  • “convertible” means that the conversion is possible without manual intervention using the controlled source language sentence DB 26.
  • the controlled source language sentence conversion possibility determination unit 35 uses a word or a combination of words included in the input source language sentence as a word included in the source language sentence stored in the controlled source language sentence DB 26 (corpus) or Compare with word combinations. Thereby, the matching degree of two sentences is calculated, and it is determined whether it can be converted into a controlled source language sentence based on the magnitude of the matching degree.
  • multilingual translation via CNL has already been put into practical use as a method for automatically translating accurate information, represented by multilingual avalanche warning.
  • converting an input source language sentence into a CNL sentence is a conversion into a source language sentence with low ambiguity. That is, even an input source language sentence determined not to be machine-translatable can be converted into a source language sentence that can be machine-translated by converting it into a controlled source language sentence. Therefore, the controlled source language sentence conversion possibility determination unit 35 determines that the input source language sentence determined to be convertible into the controlled source language sentence is machine-translatable, and converts it into the controlled source language sentence. An input source language sentence determined to be impossible is determined not to be machine-translatable.
  • the translation target data DB 27 stores a controlled source language sentence conversion possibility determination flag for each input source language sentence.
  • the DB control unit 37 updates (stores) the flag for each input source language sentence.
  • the controlled source language sentence conversion unit 36 uses the controlled source language sentence DB 26 to convert each sentence of the input source language sentence determined to be convertible into the controlled source language sentence into a controlled source language sentence.
  • the controlled source language sentence conversion unit 36 includes a controlled source language sentence conversion enable / disable determining unit 35, and determines whether or not the input source language sentence can be converted into a controlled source language sentence and converts the input source language sentence. Each sentence of the input source language sentence determined to be possible is converted into a controlled source language sentence.
  • the DB control unit 37 determines that machine translation is not possible (cannot be converted into a controlled source language sentence) among the input source language sentences in accordance with an input to the input unit 22 received by a user operation. Extract input source language sentences.
  • the output unit 23 can display the extracted input source language sentence on a display or output it to a printer.
  • the DB control unit 37 can also extract an input source language sentence determined to be machine-translatable from the input source language sentences in accordance with an input to the input unit 22 received by a user operation.
  • the input unit 22 can receive an input of a controlled source language sentence for an input source language sentence that is determined not to be convertible into a controlled source language sentence.
  • the controlled source language sentence conversion unit 36 converts the input source language sentence corresponding to the input into the controlled source language sentence that has received the input.
  • the DB control unit 37 stores the controlled source language sentence that has received the input and the corresponding input source language sentence in the controlled source language sentence DB 26.
  • the controlled source language sentence conversion availability determination unit 35 determines that the input source language sentence corresponding to the input is machine-translatable, and determines that the controlled source language sentence conversion is possible.
  • the source language sentence to be machine-translated is determined to be machine-translatable and the input source language sentence that is not determined to be controlled source language sentence conversion and the input source sentence converted to the controlled source language sentence.
  • Language sentence
  • the machine translation conversion unit 38 transmits the source language sentence to be translated into the machine translation server 6 via the communication device 15 in accordance with the input to the input unit 22 received by the user operation, and the machine translated object A language sentence is received from the machine translation server 6.
  • the machine translation conversion unit 38 stores the machine-translated target language sentence in the translation target data DB 27.
  • the output unit 23 displays the machine-language translated target language sentence on a display or outputs it to a printer in response to an input to the input unit 22 received by a user operation. At this time, the output unit 23 preferably displays or outputs the target language sentence and the corresponding input source language sentence before translation together.
  • the machine translation conversion unit 38 may perform machine translation on some source language sentences among the source language sentences to be machine-translated in response to an input to the input unit 22 received by a user operation. Good.
  • the machine translation conversion unit 38 includes a character string conversion filter unit.
  • the character string conversion filter unit replaces a predetermined word included in the source language sentence to be machine-translated with a predetermined proper noun before sending it to the machine translation server 6.
  • the character string conversion filter unit re-replaces the pre-replaced proper noun included in the target language sentence with the predetermined word before the replacement.
  • the character string conversion filter unit when receiving the machine-translated target language sentence from the machine translation server 6, sets the target language corresponding to the pre-replaced proper noun included in the target language sentence to a predetermined value before the replacement. The target language corresponding to the word may be replaced again.
  • the machine translation server 6 Since the machine translation server 6 is a general-purpose product, it does not have game proper nouns such as character names or imaginary place names as a dictionary.
  • the machine translation by the machine translation server 6 has a property that the translation accuracy is remarkably deteriorated when an unknown word appears. Therefore, the character string conversion filter unit can replace the unknown word with a general-purpose term and replace it again with the unknown word after machine translation, thereby preventing deterioration of the accuracy of machine translation.
  • FIG. 4 shows a state of filter processing by the character string conversion filter unit of the machine translation conversion unit 38 according to an embodiment of the present invention, and is an example of a screen output to the display by the output unit 23.
  • the source language is English and the target language is Japanese.
  • box 41 “Oneiros met Coux.” Is input as the source language sentence to be translated.
  • the source language sentence includes proper names of Coux and Onerios, but these are not registered in the dictionary of the existing machine translation system and cannot be translated.
  • the character string conversion filter unit converts Coux to Andy, Onerios to Bob, and common proper nouns before the source language sentence to be translated is sent to the machine translation server 6.
  • the character string conversion filter unit re-replaces Andy to Coux and Bob to Onerios, as shown in box 44.
  • the machine translation conversion unit 38 stores the machine-translated target language sentence after the replacement in the translation target data DB 27.
  • the box 42 and the box 43 are shown, but the box 42 and the box 43 are preferably not normally displayed.
  • the translation support device 10 stores the replacement candidate list in the storage device 14.
  • the replacement candidate list is associated with proper nouns such as Andy, Bob, Charles, ... for male names that are expected to be entered in the source language sentence in advance, and for female names. Is a list of proper nouns such as Anna, Becky, Carol, ...
  • the replacement candidate list includes a list corresponding to a category of proper nouns such as a person name, a weapon name, and a magic name, and the character string conversion filter unit refers to the replacement candidate list according to the category, and performs replacement. Process.
  • the replacement candidate list includes a list that converts “to is” to “to daze” etc.
  • the string conversion filter unit is The replacement process is performed by referring to the replacement candidate list according to the category.
  • the replacement candidate list is associated with proper nouns such as Andy, Bob, Charles, ..., for example, nouns in the target language such as Andy, Bob, Charles, ...
  • the character string conversion filter unit can be appropriately replaced again.
  • the machine translation server monitoring unit 39 periodically transmits a predetermined source language sentence to the machine translation server 6 (for example, once a month, once in March, or once every six months). A language sentence is received from the machine translation server 6. The machine translation server monitoring unit 39 compares the target language sentence that has been machine translated this time with the target language sentence that has been machine translated the last time (for example, one month before if machine translation is performed once a month), and based on the difference A change or update of the machine translation server 6 is detected.
  • the predetermined source language sentence is preferably a plurality of machine-translatable source language sentences.
  • step 501 the input unit 22 receives an input source language sentence.
  • step 502 the syntax filter unit 33 determines a score representing the complexity of the input source language sentence for the input source language sentence based on the sentence structure of the input source language sentence, and the magnitude of the score. To determine whether the input source language sentence can be machine translated. If it is determined that machine translation is possible, the process proceeds to step 503. If it is determined that machine translation is not possible, the process proceeds to step 504.
  • step 503 the semantic filter unit 34 determines the input source language sentence based on the word or combination of words included in the input source language sentence and the word or combination of words stored in the error DB 25. It is determined whether or not the language sentence can be machine translated. If it is determined that machine translation is possible, the process proceeds to step 506, where the machine translation availability determination unit 32 determines that the input source language sentence can be machine translated. If it is determined that machine translation is not possible, the process proceeds to step 504.
  • step 504 the controlled source language sentence conversion availability determination unit 35 determines whether the input source language sentence can be converted into a CNL sentence based on the controlled source language sentence DB 26. If it can be converted into a CNL sentence, in step 505, the controlled source language sentence conversion unit 36 converts the input source language sentence into a CNL sentence using the controlled source language sentence DB 26. Thereafter, in step 506, the machine translation availability determination unit 32 determines that the input source language sentence can be machine translated. If it cannot be converted into a CNL sentence, the machine translation availability determination unit 32 determines in step 507 that the input source language sentence is not machine-translatable.
  • the DB control unit 37 selects at least one of an input source language sentence determined not to be machine-translatable and an input source language sentence determined to be machine-translatable from a plurality of input source language sentences. Can be extracted.
  • step 601 the input unit 22 receives an input source language sentence to be machine-translated.
  • step 602 the character string conversion filter unit of the machine translation conversion unit 38 replaces a predetermined word included in the source language sentence to be translated into a predetermined proper noun.
  • step 603 the machine translation conversion unit 38 executes machine translation by transmitting the machine language translation target source language sentence subjected to the replacement process in step 602 to the machine translation server 6. A language sentence is received from the machine translation server 6.
  • Step 604 when the machine translation conversion unit 38 receives the machine-language translated target language sentence from the machine translation server 6, the machine-determined proper noun included in the target language sentence is replaced with the predetermined proper noun before the substitution. Replace with the word
  • the translation support system 1 uses the syntax filter unit 33 and the semantic filter unit 34 to determine whether the input source language sentence to be translated is machine-translatable. Judging from the viewpoint, it automatically extracts sentences that should be corrected manually to apply machine translation. As a result, it is possible to realize low-cost translation (translation support) using a machine translation system, utilizing as much source text as possible, that is, input source language text, and reducing manual text correction. Become. Further, since the machine translation server 6 that is an existing machine translation system is used, it becomes possible to realize translation (translation support) for many languages supported by the existing machine translation system.
  • the translation support system 1 can determine whether or not an input source language sentence determined not to be machine-translatable by the semantic filter unit 34 can be converted into a CNL sentence based on the controlled source language sentence DB 26. It is determined whether or not machine translation is possible depending on whether or not the conversion is possible.
  • the translation support system 1 converts each sentence of the input source language sentence determined to be convertible into a CNL sentence into a CNL sentence, and determines that machine translation is possible.
  • the translation support system 1 can machine-translate by converting an input source language sentence that is determined not to be machine-translatable into a CNL sentence by rewriting the sentence with low ambiguity. It can be a source language sentence. As a result, while maintaining high translation accuracy, it is possible to realize low-cost translation (translation support) by using as much original text as possible and reducing manual text correction.
  • the translation support system 1 can accept an input that the machine-translated target language sentence has not been properly translated, and the error DB 25 corresponds to the target language sentence that has accepted the input. Stores input source language sentences.
  • the translation support system 1 uses the error DB 25 in which the success or failure of the quality of the target language sentence actually machine-translated in this way is fed back in the semantic filter unit 34.
  • the controlled source language sentence DB 26 stores, as a pair, the result of rewriting from the input source language sentence to the CNL sentence by hand.
  • the translation support system 1 converts the input source language sentence into a CNL sentence based on the controlled source language sentence DB 26.
  • the translation support system 1 stores the DB data related to the sentence of the target game in the controlled source language sentence DB 26 and the error DB 25, so that the semantic filter unit 34 It becomes possible to improve the determination accuracy and the conversion accuracy to CNL. As a result, it is possible to maintain high translation accuracy while reducing human involvement. Also, for example, by using the translation support system 1, the content having several megabytes of text relating to the target game is translated into a plurality of languages at the cost of rewriting the entire sentence of about 30 to 50% only once. It becomes possible.
  • the translation support system 1 can be applied not only to games but also to translations of FAQs, help, websites, and chatbots.
  • the translation support system 1 performs a process for determining whether or not machine translation is possible, and then performs a machine translation server 6 that is an existing machine translation system on the source language sentence to be machine translated. Is used as a black box.
  • the translation support system 1 does not depend on a specific machine translation system.
  • the machine translation server 6 can be an arbitrary machine translation system such as Google Translate API or Microsoft Translate API. If a machine translation system developed in the future appears, it can be migrated at a very low cost. Is possible.
  • the translation support system 1 replaces a predetermined word included in the source language sentence to be translated with a predetermined proper noun before transmitting the source language sentence to be translated to the machine translation server 6. To do.
  • the translation support system 1 re-replaces the pre-replaced proper noun included in the target language sentence with the predetermined word before the replacement. For example, when translating sentences in a game, it is conceivable that many unknown words appear in comparison with general sentences. With this configuration, the translation support system 1 can prevent deterioration of the accuracy of machine translation by the machine translation server 6 due to the appearance of an unknown word.
  • the machine translation server monitoring unit 39 periodically translates a predetermined source language sentence, for example, once a month, using the machine translation server 6, and obtains a target language sentence obtained periodically. A change or update of the machine translation server 6 is detected from the difference.
  • the translation support apparatus 10 periodically monitors the translation result by the general-purpose machine translation system, and recalculates the machine translation possibility when there is a large change in the translation result. Thus, it is possible to automatically follow the update of the machine translation system.
  • FIG. 7 is an overall configuration diagram of the translation system 2 according to another embodiment of the present invention.
  • the translation system 2 includes a translation support device 10, a machine translation server 6, and an end user terminal 8.
  • the translation support apparatus 10, the machine translation server 6, and the end user terminal 8 are connected to a network 4 such as the Internet and can communicate with each other.
  • the configuration is the same as that of the translation support system 1 shown in FIG. 1 except that the translation support system 1 (translation system 2) includes the user terminal 8. Therefore, different points will be mainly described.
  • the end user terminal 8 is a terminal of a user such as an end user or a person in charge of debugging who plays the game, for example, when the machine-translated content is a game.
  • the end user terminal 8 is a computer used by a user who browses machine-translated sentences, and may be a personal computer, a tablet terminal, a smartphone, a mobile phone, or the like, for example. Since the hardware configuration of the end user terminal 8 is known, the description thereof is omitted.
  • the end user terminal 8 receives an input indicating that the machine could not understand when the machine translated sentence was viewed by the end user and the end user could not understand the meaning of the sentence (target language sentence). It is configured to be able to. For example, the end user terminal 8 displays a button such as “Report a translation problem” on the display, and accepts the input when clicked by the end user. When the end user terminal 8 accepts the input, the end user terminal 8 transmits translation feedback data associated with a sentence whose meaning could not be understood to the translation support apparatus 10.
  • the DB control unit 37 stores the input source language sentence corresponding to the target language sentence associated with the received data in the error DB 25.
  • the translation support apparatus 10 does not include at least one of the source language sentence DB 24, the error DB 25, the controlled source language sentence DB 26, and the translation target data DB 27.
  • the translation support system 1 includes a database server having a DB function that the translation support apparatus 10 does not have, and the translation support apparatus 10 is configured to be accessible to the database server.
  • a computer-readable storage medium storing a program for realizing the functions of the embodiment of the present invention described above and the information processing shown in the flowchart may be used.
  • a server capable of supplying the translation support apparatus 10 with a program that realizes the functions of the embodiment of the present invention described above and the information processing shown in the flowchart may be used.
  • the server can cause the translation support apparatus 10 to download the program via wired or wireless communication.
  • a virtual machine that realizes the functions of the embodiment of the present invention described above and the information processing shown in the flowchart may be used.
  • translation support system 1 translation support system 2 translation system 4 network 6 machine translation server 8 user terminal 10 translation support device 11 processor 12 input device 13 output device 14 storage device 15 communication device 21 control unit 22 input unit 23 output unit 24 source language sentence database (original Language sentence DB) 25 Error database (Error DB) 26 Controlled source language sentence database (Controlled source language sentence DB) 27 Translation target database (Translation target data DB) 31 Basic control unit 32 Machine translation availability determination unit 33 Syntax filter unit 34 Semantic filter unit 35 Controlled source language sentence conversion availability determination unit 36 Controlled source language sentence conversion unit 37 Database control unit (DB control unit) 38 Machine Translation Conversion Unit 39 Machine Translation Server Monitoring Unit 40 Machine Learning Units 41 to 44 Box

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Health & Medical Sciences (AREA)
  • Artificial Intelligence (AREA)
  • Audiology, Speech & Language Pathology (AREA)
  • Computational Linguistics (AREA)
  • General Health & Medical Sciences (AREA)
  • Physics & Mathematics (AREA)
  • General Engineering & Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Machine Translation (AREA)

Abstract

Le but de la présente invention est de fournir un système d'aide à la traduction qui, à l'aide d'un système de traduction automatique, peut maintenir une grande précision de traduction et réduire les coûts de traduction en diminuant la contribution en main-d'œuvre humaine. La présente invention est un système d'aide à la traduction permettant d'aider une traduction automatique d'un texte en langue source à un texte en langue cible, le système d'aide à la traduction comprenant : une unité d'entrée qui reçoit une entrée du texte en langue source à traduire ; une base de données d'erreurs qui retient au moins un mot, ou une combinaison de mots, qui sont contenus dans une pluralité de textes en langue source pour lesquels une traduction automatique du texte en langue source au texte en langue cible ne peut pas être effectuée correctement ; une base de données de texte en langue source commandé qui conserve une pluralité de textes en langue source et un texte en langue source commandé qui est un texte en langue source qui correspond à une telle pluralité de textes en langue source, qui est indiqué dans un format satisfaisant une condition prescrite, et qui est commandé ; une unité de commande qui détermine si le texte de langue source d'entrée peut être traduit automatiquement ; et une unité de sortie qui peut délivrer en sortie le texte en langue source d'entrée dont il a été déterminé qu'une traduction automatique ne pouvait pas être effectuée.
PCT/JP2018/011977 2017-03-29 2018-03-26 Système d'aide à la traduction et analogue Ceased WO2018181094A1 (fr)

Priority Applications (2)

Application Number Priority Date Filing Date Title
CN201880035286.3A CN110678868B (zh) 2017-03-29 2018-03-26 翻译支持系统、装置和方法以及计算机可读介质
US16/586,000 US11288460B2 (en) 2017-03-29 2019-09-27 Translation support system, etc

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
JP2017-065365 2017-03-29
JP2017065365A JP6227179B1 (ja) 2017-03-29 2017-03-29 翻訳支援システム等

Related Child Applications (1)

Application Number Title Priority Date Filing Date
US16/586,000 Continuation US11288460B2 (en) 2017-03-29 2019-09-27 Translation support system, etc

Publications (1)

Publication Number Publication Date
WO2018181094A1 true WO2018181094A1 (fr) 2018-10-04

Family

ID=60265759

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/JP2018/011977 Ceased WO2018181094A1 (fr) 2017-03-29 2018-03-26 Système d'aide à la traduction et analogue

Country Status (4)

Country Link
US (1) US11288460B2 (fr)
JP (1) JP6227179B1 (fr)
CN (1) CN110678868B (fr)
WO (1) WO2018181094A1 (fr)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN111160045A (zh) * 2019-12-27 2020-05-15 苏州亿歌网络科技有限公司 一种游戏资源文件翻译方法、装置及设备

Families Citing this family (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP2021177263A (ja) * 2018-07-27 2021-11-11 Ｗｏｖｎ Ｔｅｃｈｎｏｌｏｇｉｅｓ株式会社 翻訳プログラム
JPWO2021009972A1 (ja) * 2019-07-18 2021-09-13 株式会社モールサービス 自然言語処理方法、自然言語処理システム、及び、自然言語処理プログラム
CN111859998A (zh) * 2020-06-18 2020-10-30 北京百度网讯科技有限公司 篇章翻译的方法、装置、电子设备和可读存储介质
US12067370B2 (en) * 2021-06-08 2024-08-20 Sap Se Detection of abbreviation and mapping to full original term
CN114757214B (zh) * 2022-05-12 2023-01-31 北京百度网讯科技有限公司 用于优化翻译模型的样本语料的选取方法、相关装置
JP7766287B1 (ja) 2024-12-08 2025-11-10 株式会社アルティメディア Ｓｎｓ投稿システム及びプログラム

Citations (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JPS6435661A (en) * 1987-07-30 1989-02-06 Hitachi Ltd Pre-edition supporting system
JP2006004366A (ja) * 2004-06-21 2006-01-05 Advanced Telecommunication Research Institute International 機械翻訳システム及びそのためのコンピュータプログラム

Family Cites Families (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP2009075795A (ja) * 2007-09-20 2009-04-09 National Institute Of Information & Communication Technology 機械翻訳装置、機械翻訳方法、及びプログラム
US8838434B1 (en) * 2011-07-29 2014-09-16 Nuance Communications, Inc. Bootstrap call router to other languages using selected N-best translations
KR20140002171A (ko) * 2012-06-28 2014-01-08 한국전자통신연구원 자동통역방법
US9614969B2 (en) * 2014-05-27 2017-04-04 Microsoft Technology Licensing, Llc In-call translation

Patent Citations (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JPS6435661A (en) * 1987-07-30 1989-02-06 Hitachi Ltd Pre-edition supporting system
JP2006004366A (ja) * 2004-06-21 2006-01-05 Advanced Telecommunication Research Institute International 機械翻訳システム及びそのためのコンピュータプログラム

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN111160045A (zh) * 2019-12-27 2020-05-15 苏州亿歌网络科技有限公司 一种游戏资源文件翻译方法、装置及设备
CN111160045B (zh) * 2019-12-27 2021-03-23 苏州亿歌网络科技有限公司 一种游戏资源文件翻译方法、装置及设备

Also Published As

Publication number Publication date
US11288460B2 (en) 2022-03-29
CN110678868B (zh) 2023-01-20
US20200026763A1 (en) 2020-01-23
JP6227179B1 (ja) 2017-11-08
CN110678868A (zh) 2020-01-10
JP2018169715A (ja) 2018-11-01

Similar Documents

Publication Publication Date Title
JP6227179B1 (ja) 翻訳支援システム等
US10311146B2 (en) Machine translation method for performing translation between languages
US10769387B2 (en) System and method for translating chat messages
KR102025968B1 (ko) 구문 기반 사전 추출 및 번역 품질 평가 기법
US8626486B2 (en) Automatic spelling correction for machine translation
US7346487B2 (en) Method and apparatus for identifying translations
JP2021197133A (ja) 意味マッチング方法、装置、電子機器、記憶媒体及びコンピュータプログラム
US10275443B2 (en) Hybrid grammatical and ungrammatical parsing
JP2005182822A (ja) 機械翻訳の品質を評価するための方法および装置
JPH1011447A (ja) パターンに基づく翻訳方法及び翻訳システム
KR20090106937A (ko) 철자 오류 보정 시스템 및 방법
JP6817556B2 (ja) 類似文生成方法、類似文生成プログラム、類似文生成装置及び類似文生成システム
WO2015096529A1 (fr) Procédé et dispositif de traduction individualisée orientée vers un moteur de traduction automatique universelle
KR20210035721A (ko) 다중-언어 코퍼스를 이용하여 기계번역 하는 방법 및 이를 구현한 시스템
US20220058349A1 (en) Data processing method, device, and storage medium
US20150081273A1 (en) Machine translation apparatus and method
Moradshahi et al. X-RiSAWOZ: High-quality end-to-end multilingual dialogue datasets and few-shot agents
US20210312144A1 (en) Translation device, translation method, and program
CN111813978A (zh) 一种图像描述模型的生成方法、生成装置和存储介质
HK40012360B (en) Translation support system, device, and method, and computer readable medium
Ay et al. Data Augmentation with ChatGPT for Text to Sign Language Gloss Translation
HK40012360A (en) Translation support system, device, and method, and computer readable medium
JP2009157900A (ja) 対訳文抽出装置及び対訳文抽出方法
Wu et al. Fine-grained sentiment analysis with 32 dimensions
Fernandes et al. Portuguese part-of-speech tagging with large margin structure learning

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 18775992

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 18775992

Country of ref document: EP

Kind code of ref document: A1